Deep Learning-Based Japanese Speech Recognition for Robust Organizational Knowledge Capture
International Journal of Web-Based Learning and Teaching Technologies · 15.10.2026 · 21 (1) · 1-18
DOI: 10.4018/ijwltt.425050 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Web-Based Learning and Teaching Technologies, erschienen am 15.10.2026. Verfasst von Zhiyuan Li. Die Arbeit nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 3.0).
Abstract (Originalsprache)
Online Japanese teaching and organizational knowledge management depend on reliable Japanese automatic speech recognition, while existing conformer models perform poorly amid regional dialects, complex phonetics, and background noise. This paper proposes an improved conformer connectionist temporal classification framework with multi-scale acoustic feature enhancement, dialect adaptation layers, bidirectional acoustic-language model fusion, and multi-task training. Tests on 800 hours of standard, mixed-dialect, and noisy speech data and three months of real-scene deployment show the model improves dialect recognition accuracy by nearly 15% and reduces word error rate by 12% under heavy noise, with only 5 milliseconds longer inference latency than the baseline. Separate module tests show each design strengthens noise resistance, dialect adaptability, and long-sentence modeling. The framework balances accuracy and efficiency, supporting real-time oral teaching feedback and structured extraction of implicit enterprise knowledge from voice recordings.
Zitieren
APA
Li, Z. (2026). Deep Learning-Based Japanese Speech Recognition for Robust Organizational Knowledge Capture. International Journal of Web-Based Learning and Teaching Technologies, 21(1), 1-18. https://doi.org/10.4018/ijwltt.425050
Vancouver
Li Z. Deep Learning-Based Japanese Speech Recognition for Robust Organizational Knowledge Capture. International Journal of Web-Based Learning and Teaching Technologies. 2026;21(1):1-18. doi:10.4018/ijwltt.425050
BibTeX
@article{Li2026,
author = {Li, Zhiyuan},
title = {Deep Learning-Based Japanese Speech Recognition for Robust Organizational Knowledge Capture},
journal = {International Journal of Web-Based Learning and Teaching Technologies},
year = {2026},
volume = {21},
number = {1},
pages = {1--18},
doi = {10.4018/ijwltt.425050}
}
EUVIMED